Live Image Composing

Japanese

Seamless Image Cloning

We have developed a seamless image cloning algorithm extending the Poisson Image Editing.
You will see the composed image by mouse-over the image.
In the Poisson image editing, the Poisson equation is solved by an iterative Gauss-Seidel algorithm. The Gauss-Seidel algorithm is computational expensive.
On contrast, we have developed the closed form solution by modifying the Poisson equation. The proposed algorithm provides very fast seamless cloning. In addition, we can easily generate a background image by using the low-frequency component estimation.
Download the matlab code of the proposed seamless image cloning.
Please cite the following paper if you use the code.
Masayuki Tanaka, Ryo Kamio and Masatoshi Okutomi, Seamless Image Cloning by a Closed Form Solution of a Modified Poisson Problem, Proceedings of 5th ACM SIGGRAPH Conference and Exhibition on Computer Graphics and Interactive Techniques in Asia (SIGGRAPH Asia 2012), p.c08-0164-1-1, November, 2012.

In the Poisson image editing, only source image region is edited. Then, for some scenes, color of the source image is completely changed. However, the proposed algorithm optimizes the whole image region. Therefore, we can obtain natural color image. The proposed algorithm has the color consistency parameter. We can control the color consistency between the output image and the input image by adjusting the color consistency parameter.
Input image Poisson image editing Proposed algorithm
Input image Poisson image editing Proposed algorithm


Live Image Composing

We have developed the easy-to-use image composing algorithm!
The photo retouch software provides functionality of the image composing. However, the usage of the photo retouch software for the image composing is very complicated for non-experts.
Then, we have developed the image composing algorithm which allows us to compose images by just taking two shots!

Overview

Demo video

This is a demo video for the live image composing. First person takes first picture. Then, second person takes second picture. At the same time to take the second picture, the image composing is performed on the tablet. The second person can take second picture checking the composed result. Therefore, they can adjust the position by just moving the camera. Finally, the second person takes the second picture. At that time, the image composing is done!

Examples of image composing

by photo retouch software (object extraction)
Capturing time: 01:15
Image composing timeF11:04
by photo retouch software (background cloning)
Capturing time: 01:20
Image composing timeF09:08
by the proposed Live Image Composing
Capturing and Image composing time: 00:50

References

  1. Live Image Composing
    Masayuki Tanaka, Shunsuke Nakamura and Masatoshi Okutomi, Proceedings of 5th ACM SIGGRAPH Conference and Exhibition on Computer Graphics and Interactive Techniques in Asia (SIGGRAPH Asia 2012), p.c07-0163-1-1, November, 2012
  2. Seamless Image Cloning by a Closed Form Solution of a Modified Poisson Problem
    Masayuki Tanaka, Ryo Kamio and Masatoshi Okutomi, Proceedings of 5th ACM SIGGRAPH Conference and Exhibition on Computer Graphics and Interactive Techniques in Asia (SIGGRAPH Asia 2012), p.c08-0164-1-1, November, 2012

Return to Gradient-Domain Image Reconstruction